Edge detector

Author: Lingfeng Zhu

Email: jolin.windy072@gmail.com

1: The detect_edge function

Do basic edge detection. The detect_edge function find the edge and change the image by using the value 1 in each color channel, which makes the edge pixels white:

detect_edge = function(X_list, k = 1, top = "10%", color = "red") { 
  # X_list is the list matrices and k is the size of the window, 
  # top is the threshold, color is the color channel I choose
  if (color == "red") X = X_list$red
  else if (color == "green") X = X_list$green
  else if (color == "blue") X = X_list$blue
  n = dim(X)[1]
  m = dim(X)[2]
  pad_matrix = matrix(0, n+2*k, m+2*k)
  pad_matrix[(k+1):(n+k), (k+1):(m+k)] = X
  
  detect = matrix(NA, nrow = n, ncol = m)
  for (i in (k+1):(n+k)) {
    for (j in (k+1):(m+k)) {
      moving_matrix = pad_matrix[(i-k):(i+k), (j-k):(j+k)]
      detect[(i-k),(j-k)] = sd(moving_matrix)
    }
  }
  if (top == "10%") cutoff = quantile(detect, probs = 0.9)
  else if (top == "1%") cutoff = quantile(detect, probs = 0.99)
  
  X_list$red[detect >= cutoff] = 1
  X_list$green[detect >= cutoff] = 1
  X_list$blue[detect >= cutoff] = 1
  return(X_list)
}

2: Find edges in the Van Gogh painting

2.1. Find the edges in the Van Gogh painting using the function I just implemented: (Use red channel)

if (!require("png")) {
  install.packages("png")
  stopifnot(require("png"))
}

## Loading required package: png

vg = readPNG("Van_Gogh.png")

red.vg = vg[,,1]
green.vg = vg[,,2]
blue.vg = vg[,,3]

vg_list = list()
vg_list$red = red.vg
vg_list$green = green.vg
vg_list$blue = blue.vg

vg_edge = detect_edge(vg_list, k = 1, top = "10%", color = "red")


layout(matrix(1:3, ncol=3))
image(t(vg_edge$red[nrow(vg_edge$red):1,]), col = gray((1:12)/13), main="Edge of Red channel")
image(t(vg_edge$green[nrow(vg_edge$green):1,]), col = gray((1:12)/13), main="Edge of Green channel")
image(t(vg_edge$blue[nrow(vg_edge$blue):1,]), col = gray((1:12)/13), main="Edge of Blue channel")

2.2 Save the Image

n = dim(vg_list$red)[1]
m = dim(vg_list$red)[2]
vg_edge_array = array(c(vg_edge$red, vg_edge$green, vg_edge$blue), dim = c(n, m, 3))

# save the images
writePNG(vg_edge_array, "vg_edge.png")

Here is the result:

Edge of Van Gogh from Red channel:

Edge of Van Gogh from Red channel

3: Find edges in the Madison photo

3.1. Find the edges in the Madison photo using the function I just implemented: (Use all the three channels once a time)

md = readPNG("Madison.png")

red.md = md[,,1]
green.md = md[,,2]
blue.md = md[,,3]

md_list = list()
md_list$red = red.md
md_list$green = green.md
md_list$blue = blue.md

md_edge_red = detect_edge(md_list, k = 1, top = "10%", color = "red") # edge from red chnnel
md_edge_green = detect_edge(md_list, k = 1, top = "10%", color = "green") # edge from gren chnnel
md_edge_blue = detect_edge(md_list, k = 1, top = "10%", color = "blue") #  # edge from blue chnnel

3.2 Save the Images

n = dim(md_list$red)[1]
m = dim(md_list$red)[2]
md_edge_red_array = array(c(md_edge_red$red, md_edge_red$green, md_edge_red$blue), dim = c(n, m, 3))
md_edge_green_array = array(c(md_edge_green$red, md_edge_green$green, md_edge_green$blue), dim = c(n, m, 3))
md_edge_blue_array = array(c(md_edge_blue$red, md_edge_blue$green, md_edge_blue$blue), dim = c(n, m, 3))

# save the images
writePNG(md_edge_red_array, "md_edge_red.png")
writePNG(md_edge_green_array, "md_edge_green.png")
writePNG(md_edge_blue_array, "md_edge_blue.png")

Here is the result:

Edge of Madison from Red channel:

Edge of Madison form Red channel

Edge of Madison from Green channel:

Edge of Madison form Green channel

Edge of Madison from Blue channel:

Edge of Madison form Blue channel